Incidence and Types of Cardiac Arrhythmias in the Peri-Ictal Period in Patients Having a Generalized Convulsive Seizure

Abstract

Background and Objectives

Generalized convulsive seizures (GCSs) are the main risk factor of sudden unexpected death in epilepsy (SUDEP), which is likely due to peri-ictal cardiorespiratory dysfunction. The incidence of GCS-induced cardiac arrhythmias, their relationship to seizure severity markers, and their role in SUDEP physiopathology are unknown. The aim of this study was to analyze the incidence of seizure-induced cardiac arrhythmias, their association with electroclinical features and seizure severity biomarkers, as well as their specific occurrences in SUDEP cases.

Methods

This is an observational, prospective, multicenter study of patients with epilepsy aged 18 years and older with recorded GCS during inpatient video-EEG monitoring for epilepsy evaluation. Exclusion criteria were status epilepticus and an obscured video recording. We analyzed semiologic and cardiorespiratory features through video-EEG (VEEG), electrocardiogram, thoracoabdominal bands, and pulse oximetry. We investigated the presence of bradycardia, asystole, supraventricular tachyarrhythmias (SVTs), premature atrial beats, premature ventricular beats, nonsustained ventricular tachycardia (NSVT), atrial fibrillation (Afib), ventricular fibrillation (VF), atrioventricular block (AVB), exaggerated sinus arrhythmia (ESA), and exaggerated sinus arrhythmia with bradycardia (ESAWB). A board-certified cardiac electrophysiologist diagnosed and classified the arrhythmia types. Bradycardia, asystole, SVT, NSVT, Afib, VF, AVB, and ESAWB were classified as arrhythmias of interest because these were of SUDEP pathophysiology value. The main outcome was the occurrence of seizure-induced arrhythmias of interest during inpatient VEEG monitoring. Moreover, yearly follow-up was conducted to identify SUDEP cases. Binary logistic generalized estimating equations were used to determine clinical-demographic and peri-ictal variables that were predictive of the presence of seizure-induced arrhythmias of interest. The z-score test for 2 population proportions was used to test whether the proportion of seizures and patients with postconvulsive ESAWB or bradycardia differed between SUDEP cases and survivors.

Results

This study includes data from 249 patients (mean age 37.2 ± 23.5 years, 55% female) who had 455 seizures. The most common arrhythmia was ESA, with an incidence of 137 of 382 seizures (35.9%) (106/224 patients [47.3%]). There were 50 of 352 seizure-induced arrhythmias of interest (14.2%) in 41 of 204 patients (20.1%). ESAWB was the commonest in 22 of 394 seizures (5.6%) (18/225 patients [8%]), followed by SVT in 18 of 397 seizures (4.5%) (17/228 patients [7.5%]). During follow-up (48.36 ± 31.34 months), 8 SUDEPs occurred. Seizure-induced bradycardia (3.8% vs 12.5%, z = −16.66, p < 0.01) and ESAWB (6.6% vs 25%; z = −3.03, p < 0.01) were over-represented in patients who later died of SUDEP. There was no association between arrhythmias of interest and seizure severity biomarkers (p > 0.05).

Discussion

Markers of seizure severity are not related to seizure-induced arrhythmias of interest, suggesting that other factors such as occult cardiac abnormalities may be relevant for their occurrence. Seizure-induced ESAWB and bradycardia were more frequent in SUDEP cases, although this observation was based on a very limited number of SUDEP patients. Further case-control studies are needed to evaluate the yield of arrhythmias of interest along with respiratory changes as potential SUDEP biomarkers.

Introduction

Frequent generalized convulsive seizures (GCSs) in early-onset, long-standing epilepsy are a common phenotype in sudden unexpected death in epilepsy (SUDEP).^1-3 The precise pathomechanisms of SUDEP are unknown, but monitored evidence suggests a combination of cardiac rhythm and breathing dysfunction.^4,5 The role of cardiac arrhythmias in near-SUDEP and SUDEP and their relationship with GCS severity (duration and degree of oxygen desaturation, prolonged ictal central apnea [ICA], postconvulsive central apnea [PCCA], and prolonged postictal generalized EEG suppression [PGES] duration) are unknown.^6-9 We lack systematic prospective incidence data on benign and malignant cardiac arrhythmias in SUDEP. Sinus tachycardia, the most common peri-ictal arrhythmia (80% of GCS and non-GCS), is likely benign,¹⁰ whereas peri-ictal bradycardia and asystole are rare (<1% of epileptic seizures) and may not be benign.^10,11 Postictal bradycardia is commonly observed in monitored near-SUDEP and SUDEP cases.¹⁰ Rarely, dangerous arrhythmias, including atrioventricular conduction block, atrial flutter/fibrillation, and ventricular fibrillation/tachycardia,¹⁰ can occur during or after seizures and result in falls, injuries, and potential deaths. Because GCSs represent the strongest SUDEP risk factor^1,3 and almost all monitored deaths occurred after GCS,^4,12 we prospectively assessed the incidence of GCS-related cardiac arrhythmias. We also investigated the potential relationship between seizure-induced arrhythmias of interest and electroclinical GCS characteristics, including seizure severity markers.¹³

Methods

Patient Selection

All patients were prospectively consented participants in the NINDS Center for SUDEP Research's Autonomic and Imaging Biomarkers of SUDEP multicenter project (U01-NS090407) and the Prevention and Risk Identification of SUDEP Mortality Project (P20NS076965). This study was approved by the institutional review boards of all participating centers. Patients with epilepsy aged 18 years and older undergoing video-EEG (VEEG) evaluation in epilepsy monitoring units from April 2010 to October 2019 were enrolled. Inclusion criteria were patients with recorded GCS (generalized tonic-clonic seizures, focal-to-bilateral tonic-clonic seizures, and focal-onset motor bilateral clonic seizures).¹⁴ Exclusion criteria were status epilepticus and obscured video recording. Demographic and clinical data collected included sex, age, age at epilepsy onset and epilepsy duration, epilepsy type,¹⁵ GCS frequency in the year before admission, vagus nerve stimulator (VNS) therapy, peri-ictal semiologic features,¹⁶ state at seizure onset, presence of major cardiac (coronary artery disease, arrhythmia, valvulopathy) or respiratory (obstructive sleep apnea, asthma, chronic obstructive pulmonary disease [COPD]) comorbidities, and body mass index. Chronic antiseizure medication (ASM) use was classified as no therapy, monotherapy, and polytherapy. Information regarding treatment with sodium channel blockers (phenytoin, carbamazepine, oxcarbazepine, eslicarbazepine, lamotrigine, lacosamide, and rufinamide) was also collected.^17-20

Cardiorespiratory and VEEG Monitoring

Prolonged VEEG monitoring followed the 10-20 International Electrode System. Electrocardiogram (EKG), pulsioximetry, and inductance plethysmography acquisition were performed using previously described methodology.^6,7,16,21-23

We analyzed a 2-minute period before clinical or electrographic seizure onset (preictal period), the ictal period (limited to the available technically satisfactory portions of the recording), and 3 minutes after clinical seizure end (postconvulsive period). For the preictal and postconvulsive periods, if EKG data were lost or had an artifact for >6 consecutive seconds, the full period was disregarded. For the ictal period, this applied only to the nonconvulsive phase of the seizure because all of them had an unavoidable artifact during the tonic-clonic movements. Maximal heart rate (HR) and lowest HR in case of bradycardia were determined by measurement of shortest and longest RR intervals, respectively, through visual beat-to-beat analysis.

In all 3 periods, cardiac rhythm abnormalities were reviewed by a board-certified cardiac electrophysiologist (C.Y.M.). These included sinus tachycardia (>100 beats per minute [bpm]), bradycardia (<60 bpm), asystole (R-R interval ≥3 seconds), supraventricular tachyarrhythmias (atrial or junctional tachycardia [SVTs]), atrial fibrillation (Afib), premature atrial complexes (PACs), premature ventricular complexes (PVCs), nonsustained ventricular tachycardia (3 or more consecutive ventricular beats at least 20% higher than baseline sinus rhythm [NSVT]), ventricular fibrillation (VF), and atrioventricular block (first, second, or third degree [AVB]). We also noted the presence of exaggerated sinus arrhythmia (ESA) when variations in sinus rhythm were found without apparent correlation to the breathing pattern, and the presence of exaggerated sinus arrhythmia with bradycardia (ESAWB) defined by the presence of ESA with at least 1 R-R interval >1 second and <3 seconds. If the R-R interval was >1 second in more than 3 consecutive beats, this was labeled as bradycardia. If multiple arrhythmia types were recorded in the same period, all were considered in the analysis. For the purposes of this study, bradycardia, asystole, SVT, NSVT, Afib, VF, AVB, and ESAWB were classified as “arrhythmias of interest.” Arrhythmias occurring during the ictal or postconvulsive period (but not present in the preictal period) were considered seizure induced.

Breathing analysis used composite analysis of inductance plethysmography, EEG breathing artifact, and visually inspected thoracoabdominal excursions, following published methods.^7,16 Central apnea (cessation of thoracoabdominal breathing movements) was defined as 1 missed breath without other explanation (i.e., speech or intervention), with a minimum duration of 5 seconds.²⁴ ICA referred to central apnea occurring in the preconvulsive phase of GCS, and it could only be determined if thoracoabdominal belts were available.⁶ PCCA referred to central apnea after GCS and was determined either using thoracoabdominal belts or through visual inspection, given that breathing is usually stertorous and deep after GCS while the patient is immobile.⁷ Incidences and durations of ICA and PCCA were determined.

Oxygen saturation (SpO₂) was assessed using validated methods.^7,16 Baseline SpO₂ was determined as the mean value in a 15-second page at 2 minutes before EEG onset or clinical onset, whichever occurred first. We defined change in SpO₂ as the difference between baseline and the lowest SpO₂ value (nadir SpO₂) recorded up to 3 minutes after clinical seizure end. Hypoxemia was defined as SpO₂ <90%. When baseline SpO₂ was already <90%, a >1% drop was considered significant. If a transient loss of SpO₂ signal occurred during monitoring, SpO₂ nadir (and thus change in SpO₂) was not determined. Hypoxemia duration before the convulsive phase of the seizure (hypoxemia pre-GCS duration) and SpO₂ value at GCS onset (SpO₂ at GCS onset) were collected. To avoid the effect of seizure duration, following previous studies, we determined the time to hypoxemia recovery after clinical seizure end, which we termed “SpO₂ recovery.”^16,25 We considered early oxygen administration and early suction when these were applied during the seizure or within 5 seconds of seizure termination.^25,26

Presence and duration of PGES⁸ were determined by a validated automated EEG suppression detection tool²⁷ and supplemented with visual analysis by 2 epilepsy neurophysiologists (S.D.L. and L.V.) when the tool results were indeterminant.

Yearly follow-up was conducted through a combination of clinic visits, chart review, and telephone interviews. Three researchers (S.D.L., M.O.-U., and S.R.) reviewed autopsy reports (when available), circumstances of death reported in death certificates, chart reviews, and telephone interviews with the next of kin.²⁸

Statistical Analysis

Descriptive statistics for continuous variables were reported as mean ± SD and median (first quartile, third quartile). For categorical variables, number and percentage were provided.

Binary logistic generalized estimating equations (BL-GEEs) were used to determine clinical-demographic and peri-ictal variables that were predictive of the presence of seizure-induced arrhythmias of interest in those seizures that did not have arrhythmias of interest in the preictal period. The BL-GEE model was corrected for within-participant effects. A value of p < 0.05 was considered significant. Asymmetric distribution of predictive variables between the binary categories of the response variable and/or collinearity among the predictor variables are likely to cause errors in the GEE model. Under such circumstances, we eliminated the collinear variables that possessed the greater amount of missing data instead of merging the variables (a common practice to avoid Hessian singularity error). Hence, 2-staged BL-GEEs were constructed to determine seizure-induced arrhythmias of interest. The first-stage BL-GEE would help determine the significant clinical-demographic and peri-ictal seizure characteristics that were predictive of seizure-induced arrhythmias of interest. In the second-stage BL-GEE, we included these significant predictors along with peri-ictal respiratory variables, to determine the final set of predictors of seizure-induced arrhythmias of interest. This helped negate the Hessian singularity error without compromising the outcome prediction.

The z-score test for 2 population proportions was used to test whether the proportion of seizures and patients with postconvulsive ESAWB or bradycardia differed between SUDEP patients and survivors.

Data Availability

Data are available from the corresponding author, upon reasonable request.

Results

A total of 492 GCSs were reviewed in 270 patients. Of which, 455 GCSs in 249 patients (137 female [55%] with 214 seizures [53%]) met the inclusion criteria and had no exclusion criteria, with a mean of 1.8 seizures per patient (minimum 1, maximum 12). Demographic and seizure characteristics are given in Table 1 and eTable 1.

Table 1.

Demographic Characteristics per Patient and per Seizure

Variables	Patients (n = 249)	Seizures (n = 455)
Age at study, y, mean ± SD (median [IQR])	37.1 ± 13.5 (34 [26–47])	—
Age at epilepsy onset, y, mean ± SD (median [IQR])	21.6 ± 15.4 (19 [10–31])	—
Epilepsy duration, y, mean ± SD (median [IQR])	15.4 ± 12.1 (13.8 [5–23])	—
BMI, kg/m2, mean ± SD (median [IQR])	28.5 ± 6.9 (26.9 [16.4–46.9])	—
GCS frequency in the year before admission, n (%)
0	35 (16.4)	59 (14.8)
1–2	54 (25.4)	78 (19.5)
3–12	66 (31)	127 (31.8)
>12	69 (32.4)	135 (33.8)
Unknown	36	56
Cardiac comorbidities, n (%)	7 (2.9)	9 (2)
Unknown	6	6
Respiratory comorbidities, n (%)	25 (10.3)	56 (12.5)
Unknown	6	6
Epileptogenic zone, n (%)
Temporal	124 (51)	216 (48.3)
Frontal	31 (12.8)	72 (16.1)
Parietal	2 (0.8)	2 (0.4)
Occipital	1 (0.4)	1 (0.2)
Multifocal	21 (8.6)	44 (9.8)
Generalized	37 (15.2)	52 (11.6)
Lateralized	25 (10.3)	56 (12.5)
Insula	1 (0.4)	2 (0.4)
Both focal and generalized	1 (0.4)	2 (0.4)
Unknown	6	8
Neuroimaging, n (%)
Negative	112 (50.2)	211 (51.1)
Positive	111 (49.8)	202 (48.9)
Unavailable	26	42
VNS, n (%)	4 (1.6)	10 (2.3)
Unknown	6	11
ASM, n (%)
None	5 (2)	6 (1.4)
Monotherapy	79 (32.4)	120 (27)
Polytherapy	160 (65.6)	318 (71.6)
Unknown	6	11
Na channel blockers, n (%)	175 (72)	340 (76.6)
Unknown	6	11

Abbreviations: ASM = antiseizure medication; BMI = body mass index; GCS = generalized convulsive seizure; IQR = interquartile range; n = number; Na = sodium; VNS = vagus nerve stimulator.

Of note, for GCS frequency and ASM variables, the total percentage of patients adds up above 100% because some patients with more than 1 admission have had different GCS frequency and therapeutic regimens among admissions.

Frequency and Incidence of Cardiac Arrhythmias and Tachycardia

Complete preictal, ictal, and postconvulsive period EKG data were available for 397 seizures in 228 patients. Exaggerated sinus arrhythmia was the most common seizure-induced arrhythmia, with an incidence of 137 of 382 seizures (35.9%) (106/224 patients [47.3%]).

The incidence of arrhythmias of interest was 50 of 352 seizures (14.2%) in 41 of 204 patients (20.1%). Although rare, the most common seizure-induced arrhythmia of interest was ESAWB in 22 of 394 seizures (5.6%) (18/225 patients [8%]), followed by SVT in 18 of 397 seizures (4.5%) (17/228 patients [7.5%]). Further details are described below, and in Table 2 and eTables 2 and 3.

Table 2.

Frequency of Different Cardiac Arrhythmia (Arr) Types for the Different Periods and Incidence

	Preictal Arr (szs) (n = 436)	Preictal Arr (pts) (n = 241)	Ictal Arr (szs) (n = 424)	Ictal Arr (pts) (n = 237)	Ictal de novo Arr (szs) (n = 424)	Ictal de novo Arr (pts) (n = 237)	Postconvulsive Arr (szs) (n = 400)	Postconvulsive Arr (pts) (n = 229)	Postconvulsive de novo Arr (szs) (n = 400)	Postconvulsive de novo Arr (pts) (n = 229)	Incidence of Arr (szs)	Incidence of Arr (pts)
No Arr	360 (82.6)	207 (85.9)	376 (88.7)	212 (89.5)	—	—	146 (36.5)	97 (42.4)	—	—	—	—
Asystole	0	0	0	0	0	0	2 (0.5)	2 (0.9)	2 (0.5)	2 (0.9)	2/397 (0.5)	2/228 (0.9)
AVB	2 (0.5)	2 (0.8)	2 (0.5)	2 (0.8)	0	0	3 (0.8)	3 (1.3)	1 (0.3)	1 (0.4)	1/395 (0.3)	1/226 (0.4)
Bradycardia	45 (10.3)	41 (17)	17 (4)	16 (6.8)	6 (1.4)	6 (2.5)	12 (3)	9 (3.9)	7 (1.8)	4 (1.7)	13/356 (3.6)	10/208 (4.8)
ESAWB	4 (0.9)	4 (1.7)	6 (1.4)	5 (2.1)	6 (1.4)	5 (2.1)	19 (4.8)	16 (7)	16 (4)	13 (5.9)	22/394 (5.6)	18/225 (8)
SVTs	0	0	2 (0.5)	2 (0.8)	2 (0.5)	2 (0.8)	18 (4.5)	17 (7.4)	16 (4)	15 (6.5)	18/397 (4.5)	17/228 (7.5)
AFib	0	0	0	0	0	0	1 (0.3)	1 (0.4)	1 (0.3)	1 (0.4)	1/397 (0.25)	1/228 (0.4)
NSVT	0	0	0	0	0	0	3 (0.8)	3 (1.3)	3 (0.8)	3 (1.3)	3/397 (0.8)	3/228 (1.3)
VF	0	0	0	0	0	0	0	0	0	0	0	0
ESA	15 (3.4)	14 (5.8)	13 (3.1)	13 (5.5)	10 (2.4)	10 (4.2)	141 (35.3)	107 (46.7)	127 (31.8)	97 (42.4)	137/382 (35.9)	106/224 (47.3)
PAC: single	7 (1.6)	7 (2.9)	4 (0.9)	4 (1.7)	3 (0.7)	3 (1.3)	43 (10.8)	36 (15.7)	40 (10)	33 (14.4)	43/391 (11)	36/225 (16)
PAC: couplets, bigeminy, trigeminy	1 (0.2)	1 (0.4)	3 (0.7)	3 (1.3)	2 (0.5)	2 (0.8)	10 (2.5)	7 (3.1)	9 (2)	7 (2.6)	11/396 (2.7)	9/228 (3.9)
PVC: single	6 (1.4)	6 (2.5)	3 (0.7)	3 (1.3)	3 (0.7)	3 (1.3)	59 (14.8)	49 (21.4)	53 (13.3)	46 (20.1)	56/391 (14,3)	49/225 (21.2)
PVC: couplets, bigeminy, trigeminy	2 (0.46)	1 (0.4)	2 (0.5)	2 (0.8)	2 (0.5)	2 (0.8)	5 (1.3)	4 (1.7)	2 (0.5)	2 (0.9)	4/395 (1)	3/227 (1.3)

Abbreviations: AFib = atrial fibrillation; Arr = cardiac arrhythmia; AVB = atrioventricular block; ESA = exaggerated sinus arrhythmia; ESAWB = exaggerated sinus arrhythmia with bradycardia; NSVT = nonsustained ventricular tachycardia; PAC = premature auricular complex; pts = patients; PVC = premature ventricular complex; SVT = supraventricular tachyarrhythmia; szs = seizures; VF = ventricular fibrillation.

Data are presented as n (%). In bold, the most frequent Arr of interest occurring during the preictal, ictal, and postconvulsive periods, respectively, as well as the most incident Arr of interest. In italics, the most frequent Arr type occurring during preictal, ictal, and postconvulsive periods, as well as the most incident Arr.

Of note, because different arrhythmia types could be seen in the same seizure in the same period, the addition of percentages may add up above 100%.

Ictal arrhythmia occurred in 48 of 424 seizures (11.3%) (44/237 patients [18.6%]). In 4 of 424 seizures (0.9%) (4/237 [1.7%]), 2 ictal arrhythmia types were seen.

Postconvulsive arrhythmia was observed in 254 of 400 seizures (63.5%) (164/229 patients [71.6%]). Two to 4 different arrhythmias occurred in 54 of 400 seizures (21.6%) in 50 of 229 patients (21.8%). The most common combination was ESA with PVCs.

Ictal sinus tachycardia occurred in 352 of 431 seizures (81.7%) (204/262 patients [77.9%]) and postconvulsive sinus tachycardia in 422 of 428 seizures (98.6%) (238/244 patients [97.5%]).

Asystole

Postconvulsive asystole was seen in 2 of 400 seizures (0.5%) in 2 of 229 patients (0.9%). Both were monitored near-SUDEP cases, reported in a prior study.⁷ In both cases, asystole was followed by SVT. In 1 case, there was preictal bradycardia (that briefly persisted after EEG onset and was replaced by tachycardia). Asystole had a duration of 18 and 59 seconds, respectively.

Atrioventricular Block

Postconvulsive (Mobitz type I) AVB arose de novo in 1 of 400 seizures (0.3%) in 1 of 229 patients (0.4%).

Bradycardia

Ictal bradycardia occurred in 6 of 424 seizures (1.4%) in 6 of 237 patients (2.5%). Heart rate change was −40 ± 19.6 (−32 [−61 to −24]) bpm. Two patients had multiple seizures included in the study; bradycardia was recurrent in 1 patient.

Postconvulsive bradycardia occurred de novo in 7 of 400 seizures (1.8%) in 4 of 229 patients (1.7%). Change in HR from preictal to lowest HR was −23.4 ± 11 (−26 [−32 to −12]) bpm. In 2 of 3 patients with multiple seizures, postconvulsive bradycardia recurred in 2. In the other, EKG signal was lost.

Exaggerated Sinus Arrhythmia With Bradycardia

Ictal ESAWB was seen in 6 of 424 seizures (1.4%) in 5 of 237 patients (2.1%). None of the patients had preictal arrhythmias of interest, except for 1, who had preictal bradycardia. All except 1 had multiple seizures, and ictal ESAWB recurred in 1 of 4 patients (25%).

De novo postconvulsive ESAWB occurred in 16 of 400 seizures (4%) in 13 of 229 patients (5.9%). In 2 of these seizures (2 patients), preictal bradycardia was seen and 1 also had postconvulsive bradycardia. Nine patients with de novo postconvulsive ESAWB had multiple seizures, and recurrences occurred in 2 of 9 patients (22.2%).

Supraventricular Tachycardia

Ictal SVT was seen in 2 of 424 seizures (0.5%) in 2 of 237 patients (0.8%). These patients had only 1 seizure each.

De novo postconvulsive SVT occurred in 16 of 400 seizures (4%) in 15 of 229 patients (6.5%). Two were near-SUDEP patients in whom postconvulsive asystole was also seen, as described earlier in this study. Seven of the 15 patients had more than 1 seizure included in the study, and SVT recurred in 1 patient (14.3%).

Atrial Fibrillation

Nonsustained postconvulsive Afib was seen in 1 of 400 seizures (0.3%) in 1 of 229 patients (0.4%), who had only 1 seizure.

Nonsustained Ventricular Tachycardia

Postconvulsive NSVT was seen in 3 of 400 seizures (0.8%) in 3 of 229 patients (1.3%). Only 1 patient had multiple seizures and NSVT did not occur again.

Arrhythmias of Interest and Their Relationship With Electroclinical Variables

In total, 386 seizures without arrhythmias of interest in the preictal period were considered for analysis. Of these, 198 seizures (115 patients) had complete data. None of variables were associated with occurrence of seizure-induced arrhythmias of interest (Table 3).

Table 3.

Electroclinical Variables Associated With the Presence of Seizure-Induced Arrhythmias of Interest

Variable	OR	95% CI	p Value
Sex, male	3.06	0.84–11.12	0.090
Age	0.85	0.67–1.08	0.183
Age at epilepsy onset	1.18	0.93–1.50	0.168
Epilepsy duration	1.13	0.90–1.42	0.291
BMI	1.02	0.94–1.10	0.668
GCS frequency
>12	0.20	0.04–1.12	0.068
3–12	0.91	0.21–3.98	0.900
1–2	0.60	0.10–3.49	0.571
Respiratory comorbidities	0.21	0.03–1.59	0.132
Epilepsy type, generalized	0.54	0.07–3.98	0.546
Neuroimaging, positive	0.82	0.31–2.17	0.696
ASM, polytherapy	0.55	0.12–2.58	0.450
Na channel blockers	2.04	0.48–8.60	0.331
State, asleep	0.90	0.30–2.70	0.853
Tonic phase semiology
Decerebration	2.66	0.27–25.93	0.400
Decortication	2.24	0.17–29.26	0.539
Hemidecerebration	1.12	0.09–14.48	0.931
Tonic phase duration	1.12	0.98–1.28	0.098
GCS duration	1.01	0.98–1.03	0.654
Postictal posturing, yes	72.42	0.01–700,281.60	0.360
Posturing duration	0.78	0.45–1.36	0.381
Presence of PGES	0.64	0.11–3.77	0.623
PGES duration	1.02	1.00–1.05	0.060
Early O2 administration	1.46	0.52–4.06	0.471
Early suction	1.04	0.33–3.26	0.951

Abbreviations: ASM = antiseizure medication; BMI = body mass index; GCS = generalized convulsive seizure; IQR = interquartile range; n = number; Na = sodium; OR = odds ratio; O₂ = oxygen; PGES = postictal generalized electroencephalographic suppression.

When considering only those seizures in which complete respiratory data (both pulsioximetry and thoracoabdominal belts) were available, a total of 111 seizures (73 patients) were analyzed. None of the respiratory variables were associated with the presence of seizure-induced arrhythmias of interest (Table 4).

Table 4.

Respiratory Variables Associated With the Presence of Seizure-Induced Arrhythmias of Interest

Variable	OR	95% CI	p Value
Hypoxemia pre-GCS duration	0.94	0.80–1.11	0.459
SpO2 at GCS onset	1.05	0.89–1.25	0.559
SpO2 recovery	1.01	0.99–1.03	0.537
Change in SpO2	1.00	0.96–1.05	0.901
ICA	0.83	0.11–5.96	0.849
ICA duration	1.04	0.96–1.13	0.358
PCCA	1.28	0.16–10.31	0.818
PCCA duration	0.94	0.78–1.13	0.499

Abbreviations: GCS = generalized convulsive seizure; ICA = ictal central apnea; n = number; PCCA = postconvulsive central apnea; SpO₂ = oxygen saturation.

Arrhythmias of Interest and SUDEP

Eleven patients (accounting for 18 seizures) were lost to follow-up, and their vital status could not be ascertained. Five patients died of non-SUDEP–related etiologies. In a follow-up period of 48.36 ± 31.34 (43.97 [23.7–69.23]) months, there were 8 SUDEPs (2 female [2 definite, 5 probable, 1 possible]), with 18 seizures. Two had idiopathic generalized epilepsy, and the remaining had focal epilepsy. One patient with 1 seizure already had an arrhythmia of interest in the preictal period (bradycardia). Three of the remaining 7 patients (42%) had at least 1 seizure with seizure-induced arrhythmias of interest. One patient had postconvulsive SVT, 1 patient had ictal and postconvulsive bradycardia in 1 seizure and postconvulsive ESAWB in another seizure, and 1 patient had postconvulsive ESAWB. The proportion of seizures with postconvulsive ESAWB was higher in SUDEP (11%) compared with non-SUDEP (4.6%) cases (z = −12.39, p < 0.01). Accordingly, the proportion of patients with postconvulsive ESAWB was also higher in SUDEP (25%) compared with non-SUDEP (6.6%) cases (z = −3.03, p < 0.01). The proportion of non-SUDEP patients with postconvulsive bradycardia was 3.8% (3% of seizures) and that of SUDEP patients was 12.5% (5.5% of seizures) (z = −16.66, p < 0.01).

Discussion

In this large, prospective, multicenter study of cardiac arrhythmia incidence in GCS and SUDEP, we found that potentially fatal arrhythmias are rare, occurring in fewer than 1% of patients, and none are associated with electroclinical markers of GCS severity. Only 3 patients (1.3%) had NSVT, none of whom went on to die of SUDEP. By contrast, 2 of 8 SUDEP patients had transient postconvulsive bradycardia (sinus bradycardia and/or ESAWB), which deserves further investigation as a cardiac biomarker of SUDEP risk. This is in line with monitored deaths in the MORTEMUS study, where bradycardic/asystolic arrest was preceded by terminal apnea in all SUDEP patients, and some cases had transient postictal bradycardia/asystole before or simultaneously with the onset of breathing dysfunction.⁴ Thus, the role of bradycardia and ESAWB in abnormal postseizure homeostasis and SUDEP pathophysiology is of interest. Neither is likely to be the primary mechanism of death, but, together with breathing dysfunction, may produce fatal scenarios. SUDEP has been averted in animal models with strategies that restore breathing after seizures.^29-31 Given that patients have died of probable SUDEP despite well-functioning pacemakers at the time of death, the benefit of cardiac pacemakers remains unproven. Additional strategies that target breathing dysfunction may be needed to prevent death.^32,33

The fact that electroclinical seizure severity markers of PGES, hypoxemia, peri-ICA, and brainstem posturing in seizures had no consistent association with any ictal or postconvulsive seizure-induced arrhythmia of interest is interesting. This, coupled with the fact that in patents with multiple seizures, peri-ictal arrhythmias were not consistently seen in all of them, suggests that these arrhythmias are not induced by hypoxia or seizure features and may not be part of the patient's habitual seizure semiology. However, only 17% of the patients with seizure-induced arrhythmias of interest had ≥3 seizures, and hence conclusions on semiologic consistency vs stochastic phenomena cannot be inferred from this study. The thesis that hypoxemia duration during GCS^9,34,35 is arrhythmogenic is somewhat refuted by our findings, as is the theory that ictal involvement of central autonomic structures^36-39 is responsible, given the poor reproducibility of arrhythmia features in consecutive seizures. An exception to arrhythmias as a seizure phenomenon may be ictal asystole,⁴⁰ which did not occur in any of our patients. Other potential explanations include catecholamine surges, which may vary in extent between seizures and possibly cause myocardial damage. Myocardial damage is also hypothesized to occur from repeated seizure-induced ischemia; myocardial fibrosis is seen in SUDEP hearts^41,42 and, regardless of the mechanism of causation, may predispose to arrhythmias and an “epileptic heart.”⁴³ It is unlikely that ASMs played any role in generating arrhythmias given that we did not find an association between chronic treatment with sodium channel blockers and seizure-induced arrhythmias of interest. Indeed, no individual ASM has been consistently associated with an increased SUDEP risk.^18,44

In total, 82.6% of seizures had no preictal period arrhythmias in our study. Two-thirds developed ictal and postconvulsive period arrhythmias, in line with prior observations that seizures induce cardiac arrhythmias.^13,35,45,46 Seizure-induced ESA, initially described in patients with temporal lobe seizure,⁴⁷ was the most frequent arrhythmia, occurring in more than one-third of seizures. In a previous study, ESA was the second most common cardiac arrhythmia in GCS (18.8%, second only to PAC)⁴⁶ and the most common (42%) in another study of GCS and non-GCS.¹³ Sinus arrhythmia may be a surrogate of vagal tone, highly modulated by breathing changes, showing an inverse correlation to breathing rate and a direct correlation with tidal volume.⁴⁸ In a minority of seizures, ESA was associated with bradycardic beats (ESAWB). Its genesis is uncertain and may represent postseizure parasympathetic overdrive. It was over-represented in the SUDEP group as compared with the non-SUDEP group and is worthy of further study, although given the small number of SUDEP cases, this requires cautious interpretation.

The incidence of arrhythmias of interest in our prospective study largely confirmed some findings and others differed from smaller retrospective studies. The incidence of ictal/postconvulsive bradycardia was 3.6%, comparable with the 2.7% described in another study.³⁵ SVT incidence was 4.5%, similar to the 0.5%–6.3% reported previously.^35,46 Incidence of NSVT was 0.8%, ranging similarly from 0% to 2.7% in previous literature.^35,46 In accordance with prior studies, we did not see any cases of VF.^35,46 However, only 1 seizure (0.25%) in our study induced Afib, whereas other series have reported an incidence of 0%–6.3%.^35,46 Notably, ictal asystole was not observed, otherwise reported in 2.2%–6.3% of GCS series.^35,46 Our data are also comparable with a recent long-term monitoring study with implantable loop recorders that used similar criteria for defining clinically significant arrhythmias and jointly reported GCS and non-GCS.¹³ By contrast, 1 smaller long-term monitoring study that used stricter arrhythmia definitions (and excluded ictal asystole) did not detect any clinically significant arrhythmias.⁴⁹

This study has several limitations. Arrhythmia evaluation during the ictal phase could only be made during segments with readable data because of inevitable artifact on EKG channels. Thus, the presence of transient ictal arrhythmias is likely underestimated. Our definition of arrhythmias of interest was deliberately less stringent than definitions used in other studies, to capture phenomena such as ESAWB, in case these are of biomarker value in SUDEP assessments. We chose this approach because tachycardia is the norm during and after GCS, whereas postictal bradycardia has been mainly observed in (near) SUDEP cases.^10,35 Given the rarity of SUDEP, research is focused on understanding the most common responses (likely physiologic) and flagging those that are rare. We did not analyze outside the preictal, ictal, and postconvulsive periods; therefore, it is possible that some of the arrhythmias (i.e., bradycardia) were present intermittently at baseline and were not truly seizure induced. Seizure-induced cardiac arrhythmias were analyzed as a group because of the rarity of certain arrhythmia types. Even larger prospective studies designed to assess the association of seizure severity biomarkers with individual arrhythmia types and SUDEP risk may yield additional information, although we found no correlation. Another limitation is that associations between VNS treatment, cardiac comorbidities, and seizure-induced cardiac arrhythmias could not be analyzed because of insufficient number of patients in each of the groups. Finally, information about other medications potentially affecting cardiac rhythms (i.e., beta-blockers) was not available.

In conclusion, serious, potentially fatal, GCS-induced cardiac arrhythmias are rare. When they occur, markers of seizure severity appear unrelated, suggesting that other factors, such as occult cardiac abnormalities, may be relevant. Less than 7% of patients had either exclusive postconvulsive ESAWB or bradycardia, compared with 2 of 8 SUDEP patients. Our observations are based on a very limited number of SUDEP patients, and further case-control studies are needed to evaluate the yield of arrhythmias of interest along with respiratory changes as potential SUDEP biomarkers.

Glossary

Afib

atrial fibrillation

ASM

antiseizure medication

AVB

atrioventricular block

bpm

beats per minute

BL-GEE

binary logistic generalized estimating equation

EKG

electrocardiogram

ESA

exaggerated sinus arrhythmia

ESAWB

exaggerated sinus arrhythmia with bradycardia

GCS

generalized convulsive seizure

HR

heart rate

ICA

ictal central apnea

NSVT

nonsustained ventricular tachycardia

PAC

premature atrial complex

PCCA

postconvulsive central apnea

PVC

premature ventricular complex

SpO₂

oxygen saturation

SUDEP

sudden unexpected death in epilepsy

VEEG

video-EEG

VF

ventricular fibrillation

VNS

vagus nerve stimulator

Appendix. Authors

Name	Location	Contribution
Laura Vilella, MD	Departament de Medicina, Universitat Autònoma de Barcelona, Spain; NINDS Center for SUDEP Research (CSR), McGovern Medical School, University of Texas Health Science Center at Houston, Houston, TX; Department of Neurology, Hospital del Mar, Barcelona, Spain	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Christina Y. Miyake, MD	Division of Cardiology, Department of Pediatrics, Texas Children's Hospital, and Department of Molecular Physiology and Biophysics, Baylor College of Medicine, Houston, TX	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data
Ganne Chaitanya, MD, PhD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Analysis or interpretation of data
Johnson P. Hampson, MSBME	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Major role in the acquisition of data
Shirin Jamal Omidi, MD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Major role in the acquisition of data
Manuela Ochoa-Urrea, MD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Major role in the acquisition of data
Blanca Talavera, MD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Major role in the acquisition of data
Oscar Mancera, MD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Major role in the acquisition of data
Norma J. Hupp, AAS	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Major role in the acquisition of data
Jaison S. Hampson, MBBS	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Major role in the acquisition of data
M.R. Sandhya Rani, PhD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Nuria Lacuey, MD, PhD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Drafting/revision of the manuscript for content, including medical writing for content
Shiqiang Tao, PhD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Major role in the acquisition of data
Rup K. Sainju, MBBS	NINDS Center for SUDEP Research (CSR), McGovern Medical School, University of Texas Health Science Center at Houston; University of Iowa Carver College of Medicine, Iowa City	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Daniel Friedman, MD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, University of Texas Health Science Center at Houston; NYU Langone School of Medicine, New York	Drafting/revision of the manuscript for content, including medical writing for content
Maromi Nei, MD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, University of Texas Health Science Center at Houston; Sidney Kimmel Medical College, Thomas Jefferson University, Philadelphia, PA	Drafting/revision of the manuscript for content, including medical writing for content
Catherine A. Scott, MPhil	NINDS Center for SUDEP Research (CSR), McGovern Medical School, University of Texas Health Science Center at Houston; Institute of Neurology, University College London, United Kingdom	Major role in the acquisition of data
Brian Gehlbach, MD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, University of Texas Health Science Center at Houston; University of Iowa Carver College of Medicine, Iowa City	Drafting/revision of the manuscript for content, including medical writing for content
Stephan U. Schuele, MD, MPH	NINDS Center for SUDEP Research (CSR), McGovern Medical School, University of Texas Health Science Center at Houston; Department of Neurology, Feinberg School of Medicine, Northwestern University, Chicago, IL	Drafting/revision of the manuscript for content, including medical writing for content
Jennifer A. Ogren, PhD	Department of Neurobiology and the Brain Research Institute, University of California, Los Angeles	Major role in the acquisition of data
Ronald M. Harper, PhD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, University of Texas Health Science Center at Houston; Department of Neurobiology, University of California, Los Angeles	Drafting/revision of the manuscript for content, including medical writing for content
Beate Diehl, MD, PhD, FRCP	NINDS Center for SUDEP Research (CSR), McGovern Medical School, University of Texas Health Science Center at Houston; Institute of Neurology, University College London, United Kingdom	Drafting/revision of the manuscript for content, including medical writing for content
Lisa M. Bateman, MD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, University of Texas Health Science Center at Houston; Cedars-Sinai Medical Center, Los Angeles, CA	Drafting/revision of the manuscript for content, including medical writing for content
Orrin Devinsky, MD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, University of Texas Health Science Center at Houston; NYU Langone School of Medicine, New York	Drafting/revision of the manuscript for content, including medical writing for content
George B. Richerson, MD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, University of Texas Health Science Center at Houston; University of Iowa Carver College of Medicine, Iowa City	Drafting/revision of the manuscript for content, including medical writing for content
Guo-Qiang Zhang, PhD	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Major role in the acquisition of data
Samden D. Lhatoo, MD, FRCP	NINDS Center for SUDEP Research (CSR), McGovern Medical School, and Department of Neurology, McGovern Medical School, University of Texas Health Science Center at Houston	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data

Study Funding

This work was funded by NIH/National Institute for Neurological Disorders and Stroke (NINDS) U01-NS090405 and NIH/NINDS U01-NS090407.

Disclosure

L. Vilella has received honoraria as a speaker from Jazz Pharmaceuticals, Angelini Pharma, Eisai, and UCB. C.Y. Miyake is funded by NHLBI K23HL136932. D. Friedman receives salary support for consulting and clinical trial–related activities performed on behalf of the Epilepsy Study Consortium, has received research funding (paid to the Epilepsy Study Consortium) from Biohaven, BioXcell, Cerevel, Cerebral, Epilex, Equilibre, Jannsen, Lundbeck, Praxis, Puretech, Neurocrine, SK Life Science, Supernus, UCB, and Xenon, has served as a paid consultant for Neurelis Pharmaceuticals, has received travel support from the Epilepsy Foundation, has received research support (unrelated to this study) from NINDS (R01 NS109367, R01 NS233102, R01NS123928, 1U44NS121562), NSF (A20 0089 S001), and CDC (6U48DP006396), holds equity interests in Neuroview Technology, and has received royalty income from Oxford University Press. M. Nei receives support as site investigator in clinical trials funded by UCB and Eisai and receives honoraria from MedLink Neurology. S. Schuele receives grant support from NIDCD, NIMH, NINDS, and NIH, is on the speaker bureau for Neurelis, SK Life Science, Jazz pharmaceuticals, and Sunovion, and has received compensation as consultant for Epilog and Monteris. R.M. Harper receives research support from UCLA Innovation funds (no personal income received for these activities), and holds patents for the use of devices to treat migraine and breathing disorders (patents owned by UCLA). O. Devinsky receives grant support from NINDS, NIMH, MURI, CDC, and NSF, has equity and/or has received compensation from the following companies: Tilray, Receptor Life Sciences, Qstate Biosciences, Hitch Biosciences, Tevard Biosciences, Regel Biosciences, Script Biosciences, Actio Biosciences, Empatica, SilverSpike, and California Cannabis Enterprises (CCE), has received consulting fees from Zogenix, Ultragenyx, BridgeBio, GeneMedicine, and Marinus, holds patents for the use of cannabidiol in treating neurologic disorders (patents owned by GW Pharmaceuticals), holds other patents in molecular biology, and is the managing partner of PhiFund Ventures. G.B. Richerson, Guo-Qiang Zhang, and S.D. Lhatoo are funded by NINDS/NIH. S.D. Lhatoo has served on the advisory board for UCB Pharma and LivaNova Inc. The other authors report no relevant disclosures. Go to Neurology.org/N for full disclosures.